A Destination Prediction Model based on historical data, contextual knowledge and spatial conceptual maps

Mobile Wireless Network technology has enabled the development of increasingly diverse applications and devices resulting in an exponential growth in usage and services. One challenge in mobility management is the movement prediction. Prediction of the user's longer-term movement (e.g., 10 min in advance) with reasonable accuracy is very important to a broad range of services. To cope with this challenge, this paper proposes a new method to estimate a user's future destination, called Destination Prediction Model (DPM). This method combines two types of approaches: one based on the use of filtered historical movement pattern and another based on contextual knowledge; both approaches use spatial conceptual maps. The filter is based on the day and the time of the day to increase accuracy. The current movement direction, that takes into account the recent data, is used by the proposed method to reduce historical and contextual knowledge mistakes. Simulations are conducted using real-life data to evaluate the performance of the proposed model. For subjects with low predictability degree, DPM reaches an average prediction accuracy of 79%; it reaches 91% for subjects with high predictability and 86% for other subjects. Simulation results also indicate that DPM significantly reduces the impact of learning period and the remaining distance to reach the destination on prediction performance. In the future, we plan to extend our research work by proposing a full Path Prediction Model (PPM) based on the Destination Prediction Model (DPM).